基于近红外光谱技术的稻米蛋白质含量快速无损检测  

Rapid and Non-destructive Detection of Rice Protein Content Based on Near Infrared Spectroscopy

作　　者：谭思萍 岳纪成 陈莹[1] 黄翠红[1] 周丹华[1] 张惠娟 杨瑰丽[1] 王慧[1] TAN Siping;YUE Jicheng;CHEN Ying;HUANG Cuihong;ZHOU Danhua;ZHANG Huijuan;YANG Guili;WANG Hui

机构地区：[1]华南农业大学国家植物航天育种工程技术研究中心,广东广州510642

出　　处：《广东农业科学》2024年第10期111-123,共13页Guangdong Agricultural Sciences

基　　金：国家重点研发计划项目(2022YFD1200703);广东省重点领域研发计划项目(2022B0202060006);国家水稻产业技术体系专项(CARS-01-18);广州市科技计划项目(2024E04J0172)。

摘　　要：【目的】构建水稻籽粒蛋白质含量快速无损检测技术。【方法】采用317份水稻种质资源,基于近红外光谱技术(NIRS),采用一阶平滑求导(SG1)、二阶平滑求导(SG2)、标准正态变量(SNV)、去趋势算法(Detrend)4种预处理方法,结合偏最小二乘法(PLS)建立稻谷、糙米、精米3种不同形态稻米蛋白质含量近红外检测模型。【结果】预处理方式以复合式的SG1+SNV+Detrend、SNV+Detrend+SG1建模效果最佳。用于测定糙米蛋白质含量的稻谷、糙米近红外检测模型的校正相关系数(R2)分别为0.882、0.926,标准偏差(SEP)分别为0.239、0.213;测定精米蛋白质含量的稻谷、精米近红外检测模型的校正R2分别为0.900、0.925,SEP分别为0.267、0.224。测定糙米蛋白质含量的稻谷、糙米近红外检测模型的内部交叉验证相关系数(R2)分别为0.859、0.917,内部交叉验证标准偏差(SECV)分别是0.266、0.226;测定精米蛋白质含量的稻谷、精米近红外检测模型的内部交叉验证R^(2)分别为0.880、0.916、SECV分别是0.296、0.238。测定糙米蛋白质含量的稻谷、糙米近红外检测模型的外部验证相关系数(R2)分别为0.902、0.923、外部验证SEP分别为0.422、0.311;测定精米蛋白质含量的稻谷、精米近红外检测模型的外部验证R2分别为0.950、0.981,外部验证SEP分别为0.364、0.197。【结论】基于近红外光谱技术所构建的稻米蛋白质含量预测模型可用于大样本育种材料的蛋白质含量初筛,为稻米营养品质育种提供参考。【Objective】It aims to establish a non-destructive and rapid screening technique for rice kernel protein.【Method】A total of 317 rice germplasm resources were selected.Based on near infrared spectroscopy(NIRS),four pretreatment methods were used:first-order smooth derivative(SG1),second-order smooth derivative(SG2),standard normal variable(SNV)and detrend algorithm(Detrend).The near infrared detection model of rice protein contents in rice,brown rice and milled rice were established by using partial least square(PLS)method.【Result】Combined SG1+SNV+Detrend and SNV+Detrend+SG1 pretreatment methods had the best modeling effect.The corrected correlation coefficients(R2)and standard deviations(SEP)of the near infrared detection model for determination of protein content in rice and brown rice were 0.882 and 0.926,and 0.239 and 0.213,respectively.The calibration R2 and SEP of NIR detection models for determination of protein content in rice and milled rice were 0.900 and 0.925,and 0.267 and 0.224,respectively.The internal cross-validation correlation coefficients(R2)and internal cross-validation standard deviations(SECV)of NIR detection models for determination of protein content in rice and brown rice were 0.859 and 0.917,and 0.266 and 0.227,respectively.The internal cross-validation R^(2) and SECV of NIR detection models for determination of protein content in rice and milled rice were 0.880 and 0.916,and 0.296 and 0.238,respectively.The external validation correlation coefficients(R2)and external validation SEP of NIR detection models for determination of protein content in rice and brown rice were 0.902 and 0.923,and 0.422 and 0.311,respectively.The external validation R2 and external validation SEP of NIR detection models for determination of protein content in rice and milled rice were 0.950 and 0.981,and 0.364 and 0.197,respectively.【Conclusion】The prediction model of rice protein content based on near infrared spectroscopy can be used for the preliminary screening of protein content in large sampl

关 键 词：稻米 近红外光谱 蛋白质含量 近红外检测模型 偏最小二乘法 相关系数 

分 类 号：S330[农业科学—作物遗传育种]